A PCCN-Based Centered Deep Learning Process for Segmentation of Spine and Heart: Image Deep Learning

A PCCN-Based Centered Deep Learning Process for Segmentation of Spine and Heart: Image Deep Learning

K. Uday Kiran, Gowtham Mamidisetti, Chandra shaker Pittala, V. Vijay, Rajeev Ratna Vallabhuni
DOI: 10.4018/978-1-7998-9640-1.ch002
(Individual Chapters)
No Current Special Offers


The spinal cord and heart in the body are major organs. Diagnosis of diseases in these organs is very complex using MRI and CT images. The conventional methods like post segmentation, pre-image processing, and text feature extraction mechanisms cannot handle accurate diagnosis. Therefore, advanced techniques are needed. In this work, pixel-based convolution neural networks with centered deep learning processes are proposed to cross over the problems. The projected PCNN has four pixel-based convolution neural networks. Here disease objects are identified through grading framework. The entire mechanism is working based on sequential part of PCNN segmentation process. The spinal cord and heart image MRI-based diagnosis process is very difficult with conventional methods. But the proposed method provides accurate results and outperforms the standard methodology performance measures in accuracy, precision, and F1score.
Chapter Preview


The rapid growth in deep learning technology it can provide extreme results in image processing and any application specific feature extraction. Especially MRI medical image features are extracted through lower, front, up and down view disease diagnosis is very critical. These problems are overcome by proposed PCNN deep learning technology. The LBP and HBP mechanisms are very useful and creating pixel-based repair in processing time. It is usually very complex through old conventional methods. The nature of MRI image features extraction and disease finding is very simple through proposed PCNN process. The behaviors of abnormal conditions in MRI, heart and spine treatment are very difficult to track the exact damaged area. This can be very easy through proposed PCNN deep learning mechanism. The multi scale and slices segmentation process is extracting the pixels, where disease is located (Ahammad, S. H., 2020). The SVM model was trained using 1,040 pictures from 26 pregnant women and evaluated with 800 photographs from a different group of 24% women. The suggested approach has a successfulness of 97.12% on the training dataset and 94.2% on the test set. In 45 of the instances, the trained support vector machine model correctly identified the appropriate needles injection location (intervertebral area) using 46 off-line recorded Scans (Ahammad, S. H.2019).

The continuous monitorization of MRI medical image process providing an accurate observation result. But this functionality is very difficult by using machines and humans. Therefore, a deep learning mechanism is called to promote above complex work to make easy. In this impartial addition to marketing to mind when you think of any picture classification or image segmentation. Maybe something more along the lines of: provide a 512x512 picture as an input and receive a 10x1 classifier result. In the middle, there are convolution and max-pooling layers, which gradually reduce picture resolution yet preserving essential characteristics. Rather than using a traditional neural net like the one described above, Google Research and Brain Team developed SpineNet at CVPR-2020. This network, which comprises of scale-permuted intermediate features and bridge, may be a viable solution for a variety of extracting features purposes such as object recognition and picture classifications (Ahammad, S. K., 2018) .

In this discussion the calculations of spinal cord extraction very difficult in shape or appearance point of view. The adjusted conventional models can't suggest by researchers and scientist due to performance and accuracy point of view. The past mechanisms and their applications are majorly depending on old programming languages, those are not that much robust. The programming methods like python and spring java providing accurate image training mechanism and easily getting disease diagnosis area as well as location(Ahammad, S. H., 2019)..

Complete Chapter List

Search this Book: